See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/265425113

Sidescan sonar image processing: Correcting brightness variation and

patching gaps

Article in Journal of Marine Science and Technology · December 2010

DOI: 10.51400/2709-6998.1935

CITATIONS READS

42 6,940

3 authors, including:

Shu-Kun Hsu Ching-Hui Tsai

National Central University National Central University
223 PUBLICATIONS 6,089 CITATIONS 70 PUBLICATIONS 1,234 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Shu-Kun Hsu on 18 March 2016.

The user has requested enhancement of the downloaded file.

Journal of Marine Science and Technology, Vol. 18, No. 6, pp. 785-789 (2010) 785

SIDESCAN SONAR IMAGE PROCESSING:

CORRECTING BRIGHTNESS VARIATION
AND PATCHING GAPS

Yet-Chung Chang*, Shu-Kun Hsu**, and Ching-Hui Tsai**

Key words: sidescan, sonar, brightness, gap patching, mosaic, foot- from a point just below the towfish to a limited distance away
print. from the line of travel on both sides. The raw sonar data are
two time series of digitized sound: the back-scattered signals
from each ping on the port and starboard sides of the towfish.
ABSTRACT
If we display the data as gray level pixels representing the
Sidescan sonar data are typically presented as gray level acoustic energy, the result is a image similar to the left side of
images. However, sidescan images often show striking varia- Fig. 1. This is called a waterfall display.
tions in brightness. These variations, caused by the sonar Images based directly on the raw time series data have a
beam pattern and the constantly changing attitude of the tow- blind zone in the center, so for most purposes, we do not use
fish, make the images difficult to read as pictures of the seabed. the data in this raw form. The raw waterfall needs to be proc-
This reduces the utility of the images for marine geologists. essed into an image which roughly corresponds to a plane area
In this paper, the effects of the sonar beam pattern across the of the sea floor. The most basic form of processing for sides-
swath are corrected using a normalization process based on the can data is “slant range correction” [5, 6, 8]. For this process,
average signal intensity for each grazing angle. Noise along we assume that the seabed is a perfect flat plane. Then, given
the track, which is generally caused by changes in the attitude the altitude of the towfish, and the time at which the back-
of the towfish, is eliminated using an assumption that the total scatter reaches the sonar array, we can calculate a position on
back-scattered energy from each ping should be similar to the seabed. After processing the data, the blind zone in the
adjacent pings in the time series. center of image disappears, and every datum is relocated to a
These two processes allow the production of smooth and position more representative of the actual seabed. Figure 1
clear waterfall displays. However, when the sidescan data shows an example of a waterfall image before and after slant
are projected onto a map as a high resolution “mosaic” image, range correction.
gaps appear on the outer edge of the swath where the towfish
changes course. To remove the gaps, a patching algorithm is
proposed. The patching method is justified theoretically by
the shape of the sonar footprint on the sea floor. All the pixels
falling inside the sonar footprint can be assigned reasonable
values using just the original data series, with no need for
interpolation.

I. BASIC PROCESSING
The sidescan sonar is a powerful, versatile but low cost tool
Sonar track

for surveying the sea floor [1-5, 7]. Usually a ship tows a
towfish mounted with two sonar arrays, one on each side. The
sonar arrays emit fan-shaped sonar signals perpendicular to
the direction of travel. The signals scan a swath of sea floor

Paper submitted 05/22/09; revised 12/14/09; accepted 01/08/10. Author for

correspondence: Yet-Chung Chang (e-mail: [email protected]).
*Department of Digital Entertainment and Game Design, Taiwan Shoufu Fig. 1. Sidescan data before and after slant range correction. The scan
University, Tainan County, Taiwan, R.O.C. range for each side is about 400 m. The terrain features shown
**Department of Earth Sciences, National Central University, Taiwan, R.O.C. are several elongated small normal faults on a slope of sea bottom.
786 Journal of Marine Science and Technology, Vol. 18, No. 6 (2010)

Fig. 2. A typical beam pattern for sidescan sonar equipment [6]. The
solid curves represent the relative intensity of sonar emission in
different directions.
Fig. 4. Images before and after beam pattern correction.

Since the sidescan is actually two-sided sonar, each side has

its own main lobe pointed outward from the vertical line about
45 degrees, the maximum intensity therefore located around
this angle, rather than right below the towfish (see Fig. 2).
To compensate for the uneven illumination, we need to know
the energy distribution function relative to the angle. One sim-
ple way to find the energy distribution is to sum up the energy
levels for each angle over the whole data series. The result is
a histogram like Fig. 3. Using this statistical result, we can
-90 0 90 calculate the average energy for each angle; the inverse of this
average can be applied as a correcting factor to individual data
Fig. 3. Back-scattered energy as a function of grazing angle. The hori-
zontal axis denotes the grazing angle; the vertical axis denotes the in the time series. The images before and after this correction
relative intensity. are shown in Fig. 4. The unwanted variation across the swath
has been successfully removed.
Most processing procedures for the same image correction
All the images in this paper were processed based on 16-bit purpose are executed by time variable gains (TVG), either at
raw intensity data, where the darker gray color means higher source or in post-processing. However, the time based func-
intensity of backscattering. In general, the 16-bit raw data tions are not suitable to describe the variation of backscat-
allows more flexible contrast exaggeration processing than the tering energy which is basically controlled by the grazing angle.
8-bit data set. Most images in this paper had been enhanced in For example, the TVG’s must be continuously adjusted to
gray level contrast to some extent without any side effect of adapt the depth changes. In the other hand, the proposed
color discontinuity. If the case was 8-bit, many color discon- method based on grazing angle in this paper would be nearly
tinuity or bad quality within the images would be inevitable. invariant with the depth. It seems to be a more convenient and
theoretically proper method for most data processors.
Since the function of beam pattern here was established by
II. BEAM PATTERN CORRECTION
summing up the data set itself, it was possible to rule out some
The most prominent feature of sidescan images like Fig. 1 real signals right at certain grazing angle by this method. How-
is the too strong intensity near the path of the towfish, and the ever, the possibility of certain signals always appear around
weak response at the outer edge of the swath. This is caused certain grazing angle is rare. In most cases, the method has
by the beam pattern of the sonar [8]. A typical sidescan sonar provided good performance.
beam pattern is shown in Fig. 2. Across the swath, the amount
of sonar energy hitting the seabed will vary with the distance
from the towfish. The angle at which the sonar wave hits the
III. PING ENERGY LEVEL NORMALIZATION
sea floor (grazing angle) also varies. The raw sidescan sonar The energy broadcast by the sonar array should be exactly
data are the back-scattered energy from the seabed; both the the same for each ping. However, as the towfish is pulled
level of incident energy and grazing angle will affect the data. under the water, it is constantly moving because of waves and
 Y.-C. Chang et al.: Sidescan Sonar Image Processing: Correcting Brightness Variation and Patching Gaps 787

Fig. 5. Total ping energy levels. The horizontal axis denotes the series
pings and the vertical axis denotes the average intensity of each
ping.

Fig. 7. An example of mosaic image with many gaps.

Moreover, we should keep in mind that the number of “pings”

is not an independent variable proportional to the real space
occupation. The surveying speed and footprint’s width are also
involved, which makes the criteria for reference window defi-
nition quite complicate. So far, we can only adjust the aver-
aging window according to the sense of geology knowledge
subjectively.

Fig. 6. Images before and after the ping energy level normalization.
IV. PATCHING GAPS IN MOSAIC IMAGES
The purpose of a sidescan sonar survey is to determine the
currents. This causes the energy actually incident on the sea geographical location of targets or seabed features. The sonar
floor to vary from ping to ping. For example, rolling motion data therefore need to be assembled into a “mosaic” which
might cause the energy level to rise on one side and drop on represents the geophysical features of the seabed. The data for
the other side. The level of back-scattered energy also changes each ping are projected as in a waterfall display, but the ping
with the surface character of the seabed. So the energy levels lines are placed on a map, perpendicular to the path of the
can change abruptly between pings, and these abrupt changes towfish. Because this path is not always straight, the projected
show up as the white and black horizontal lines in Fig. 4. line segments are not all parallel to each other (unlike a water-
To suppress this kind of noise, we assume that each ping fall display). The result can be seen in Fig. 7 gaps appear on
“should” have a similar energy level to the pings before and the outer edge of the swath where it curves.
after it. The total back-scattered energy level for each ping is The gaps are certainly not a desirable feature in our image.
calculated, and represented as the histogram in Fig. 5. For Several processes have been suggested to deal with them, such
each ping, we find the average energy level of the 20 pings as interpolation, or repetition of the surrounding data to fill the
before and after it. This average is used as a reference value, empty pixels [6, 8]. These processes can successfully smooth
and the total energy of the ping is smoothed to match it. The the image. However, no theoretical justification has been pro-
result is that the abrupt signal changes are mostly removed. duced for why the gaps should be patched.
The images before and after this process are shown in Fig. 6. One of the assumptions in the mosaicking process is that
It should be mentioned that the assumption here does not the sonar array scans only a “line segment”. However, this
always hold. The ping energy level does vary in reality when assumption is untrue. The sonar signal has a finite width, and
the character of the sea floor changes. That is why the smooth- it spreads as it travels. Therefore the footprint of one ping is
ing must be done with reference to a fairly short sequence of actually a fan-shaped segment rather than a line, as shown in
pings (40 in this example). If the smoothing is too extensive, Fig. 8.
real changes in the sea floor will be obscured. As Fig. 8 shows, the area of the swath furthest from the
788 Journal of Marine Science and Technology, Vol. 18, No. 6 (2010)

Fig. 8. Fan-shaped sonar footprints. Horizontal axis denotes the track of

sonar moving; the vertical axis denotes the across track footprints.

Outside
a footprint Inside
O
b
Patch point

Fig. 9. Patching method based on the fan-shaped footprint.

Fig. 10. Mosaic image patched using the footprint process.

towfish is actually scanned more than areas near the path of
the towfish, because of the way the sonar beam spreads out. In
a mosaic, therefore this area should not be blank. There should
not be more gaps than in the middle of the swath. So in most
cases, the gaps should be patched not only because they are
bad for image continuity, but also because they are theoreti-
cally unjustified.
Figure 9 shows a patching method based on a fan-shaped
sonar footprint. The mosaic image is first plotted in the tradi-
tional way, with all the data projected as line segments. After
that, all empty pixels on the image are checked to find if they
are within the footprint of any sonar ping. If they are not, no
patching will be carried out. If they are, the nearest ping (de-
fined by the perpendicular distance to the line segment) is
selected as the data source. The empty pixels are assigned a
raw intensity level from the data for that ping, based on their Fig. 11. Mosaic image patched by interpolation algorithm [6]. The
distance from the towfish (O). In Fig. 9, the empty pixel “a” is most across-track gray lines overlay the images are patched
assigned the same intensity as b. When this patching method lines.
is applied to Fig. 7, the result is as shown in Fig. 10. The con-
tinuity seems perfect.
When the pixel is actually covered by more than one fan- for sidescan sonar data. Two of the processes address the un-
shaped footprint, it seems reasonable that we can introduce even brightness inherent in the sonar data. One is to patch the
some weighting algorithm to combine the contributions of all gaps in mosaic images. The three processes are listed below:
pings into the pixel’s intensity. However, it usually causes
some smoothing effect, which makes the patched lines them- 1. Processing to compensate the beam pattern (angle and in-
selves become false features. An example of similar situation tensity) varies across the swath.
is shown in Fig. 11. The main reason for this is the back- 2. Normalizing ping energy levels to remove the effects of sud-
ground intensities of sidescan images are generally with high den energy level changes, caused largely by towfish changes
frequency noises. Therefore choose only one raw datum for in attitude.
each pixel might be a better solution. 3. Patching gaps using a method supported theoretically by
the shape of the sonar footprint.
V. CONCLUSION
All of these processes are successfully applied in our cur-
This paper describes three important processing procedures rent projects.
 Y.-C. Chang et al.: Sidescan Sonar Image Processing: Correcting Brightness Variation and Patching Gaps 789

REFERENCES agement of the Monterey Bay digital sidescan-sonar images,” Marine Ge-
ology, Vol. 181, pp. 305-315 (2002).
1. Blondel, Ph., “Automatic mine detection by textural analysis of COTS 5. Fish, J. P. and Carr, H. A., Sound Underwater Images- A Guide to the
sidescan sonar imagery,” International Journal of Remote Sensing, Vol. 21, Generation and Interpretation of Side Scan Sonar Data, American Under-
No. 16, pp. 3115-3128 (2000). water Search and Survey, Lower Cape Publishing, Orleans, MA, 188p
2. Brown, C. J., Cooper, K. M., Meadows, W. J., Limpenny, D. S., and Rees,
(1990).
H. L., “Small-scale mapping of sea-bed assemblages in the Eastern English
6. Hsueh, D. Y., Development of Sidescan Sonar Image Mosaicing Software,
channel using sidescan sonar and remote sampling techniques,” Estuarine,
Master Thesis, Institute Of Applied Marine Physics and Undersea Tech-
Coastal and Shelf Science, Vol. 54, pp. 263-278 (2002).
3. Charlot, D., Girault, R., and Zerr, B., “DELPH-SONAR: A compact system nology, National Sun Yat-sen University, Kaohsiung, Taiwan (2007).
for the acquisition and processing of side scan sonar images,” OCEANS’94., 7. Huvenne, V. A. I., Blondel, Ph., and Henriet, J.-P., “Textural analyses of
Oceans Engineering for Today’s Technology and Tomorrow’s Preservation sidescan sonar imagery from two mound provinces in the Porcupine
Proceedings, Vol. 2, pp. II/428-432 (1994). seabight,” Marine Geology., Vol. 189, pp. 323-341 (2002).
4. Chavez, P. S., Jr., Isbrecht, J., Galanis, P., Gabel, G. L., Sides, S. C., Soltesz, 8. Johnson, H. P. and Helferty, M., “The interpretation of side-scan sonar,”
D. L., Ross, S. L., and Velasco, M. G., “Processing, mosaicking and man- Reviews of Geophysics, Vol. 28, No. 4, pp. 357-380 (1990).

View publication stats